OpenAI Unveils Its First Custom Chip, Built by Broadcom

My take: For years OpenAI depended on Nvidia chips and nobody knew if that would ever change. Jalapeño is the answer: an inference processor (the chip that runs the model when you talk to it) designed from scratch with Broadcom, aiming to cut the cost of running models in half. But what really left me thinking is not the chip, it is the timeline: nine months from design to tape-out, accelerated partly by OpenAI's own models. In other words, AI helping build the hardware that makes it run, like a recipe that cooks itself. If that scales, AI hardware is going to iterate far faster than the industry expected, and depending on a single vendor starts to crack. Sure, that is my opinion.